Hue-Bond has already pointed out to the problems with your code, and given a working version.

use warnings; use strict;

open my $INF, "<", "look.txt" or die "cannot open";
my @look = <$INF>;
close($INF);

my @out = sort { (split /\|/, $a)[4] <=> (split /\|/, $b)[5] } @look;

for my $line (@out) {
        my @fields = split /\|/, $line;
        print join(",", @fields[0 .. 4]), "\n";
}

__END__
[download]

use warnings; use strict;

open my $INF, "<", "look.txt" or die "cannot open";
my @look = <$INF>;
close($INF);

my @key = map { (split /\|/)[4] } @look;
my @out = @look[sort { $key[$a] <=> $key[$b] } 0 .. @look - 1];

for my $line (@out) {
        my @fields = split /\|/, $line;
        print join(",", @fields[0 .. 4]), "\n"; 
}

__END__
[download]

I'll also show you a ruby solution as a teaser.

#!ruby -w

look = File.open("look.txt").readlines; # here I'm inclined to write '
+map' instead of 'readlines'

out = look.sort_by {|line| line.split("|")[4] };

out.each {|line|
        fields = line.split "|";
        puts fields[0 .. 4].join(",");
}

__END__
[download]

That's it. You don't have to listen to the community's whining about strictures, or declare variables, nor do you have to learn complicated idioms for efficently sorting by a field, or quote regexp metacharacters if the language can do that for you. :)